Xen devel - Oct 2006 - 32bit HVM PV drivers on 64bit hypervisor

Hi Steven,
I onced played with your PV-on-HVM driver before it is checked in into
unstable, I remember at that time, there is a XENFEAT_64bit_shared_info
flag to handle situation of 32bit guest on 64bit hypervisor, at least
vnif works fine with it.

Later, when the code is checked in, this flag is gone. At first I though
this was an effort to make hypervisor more transparent to HVM guest, but
now I found 32bit guest vnif refused to work on a 64bit hypervisor. It
seems this part of code has never made its way into unstable?

Were there any considerations on this issue?

Best regards,
Qing

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 26/10/06 11:16, "He, Qing" <qing.he@intel.com> wrote:
> I onced played with your PV-on-HVM driver before it is checked in into
> unstable, I remember at that time, there is a XENFEAT_64bit_shared_info
> flag to handle situation of 32bit guest on 64bit hypervisor, at least
> vnif works fine with it.
> 
> Later, when the code is checked in, this flag is gone. At first I though
> this was an effort to make hypervisor more transparent to HVM guest, but
> now I found 32bit guest vnif refused to work on a 64bit hypervisor. It
> seems this part of code has never made its way into unstable?
This will be fixed as part of the 32-on-64 work for getting PAE PV guests
running on 64b hypervisor. This way the different format will be handled
entirely within Xen with no guest-visible interface changes.

 -- Keir



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

> I onced played with your PV-on-HVM driver before it is checked in into
> unstable, I remember at that time, there is a XENFEAT_64bit_shared_info
> flag to handle situation of 32bit guest on 64bit hypervisor, at least
> vnif works fine with it.
> 
> Later, when the code is checked in, this flag is gone. At first I though
> this was an effort to make hypervisor more transparent to HVM guest, but
> now I found 32bit guest vnif refused to work on a 64bit hypervisor. It
> seems this part of code has never made its way into unstable?
You''re right: this worked in the pre-checkin version of the PV-on-HVM
drivers, but doesn''t work in current unstable.  When the rest of the
work was checked in, the consensus was that a guest-visible feature
flag probably wasn''t the right way of handling the ABI differences,
and so that part was left behind.

I think the hope is that 32-on-64 HVM mode will come for free once
32-on-64 PV mode is working, but I''m afraid I haven''t been
following
this work particularly closely.

Steven.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel